Context

This project aims at exploring how various metrics/proxies can give insights on the vertical distribution of the phytoplankton biomass in the water column using the data measured during the Green Edge oceanographic cruise conducted in the Baffin Bay in 2016. Four main indicators were considered to get insights about the phytoplankton biomass in the water column:

  1. Fluorescence
  2. Pigments
  3. Particle beam attenuation coefficient (CP)
  4. Particle backscattering coefficient (bbp)

Most of the data is presented as a function of open water days (OWD) that were calculated in Randelhoff et al. (2019). The first graph shows the OWD of each station.

A note on the CTD data

After exploring the CTD data, I found out that there were some unwanted peaks in both the fluorescence and transmittance data. Therefore, I have used a median moving median window (\(n = 25\)) to smooth the data (both transmittance and fluorescence).

Data processing for 3D plots

During the meeting we had on 2021-01-22, it was questioned why I was averaging data by owd and depth before doing the interpolation. This is simply because there may be many observations for a specific pair of owd and depth. For example, it can be seen here that there are 9 values (i.e. 9 stations) that have an owd = 2 and depth = 1.980.

owd depth_m N
2 1.980 9
2 2.178 9
2 2.376 9
2 2.574 9
2 2.772 9
2 2.970 9

Here are 9 stations in the CTD data with owd = 2 and depth = 1.980.

Hence, before performing the interpolation, data has to be averaged by OWD and depth.

Fluorescence

This section shows the vertical distribution of phytoplankton based on fluorescence.

Fluorescence from the CTD

Using boxplots can be one interesting alternative way to present the data. For example, I have divided the CTD fluorescence data presented in the above graphic into ice-free/ice-covered and above/below the 0.1 isolume.

Fluorescence from the MVP

Note that because the MVP is measuring continually, there are no stations associated with each measurement. This is why that the MVP data is not presented as a function of OWD.

Pigments

These graphs were made using the pigments data from the rosette. Pigments were summed into two groups:

  1. Photoprotection pigments
  2. Accessory pigments

Particle beam attenuation coefficient (CP)

The increase of cp during the lighted portion of the day is usually explained by the accumulation of intracellular carbon concentration associated with photosynthetic processes (Kheireddine and Antoine 2014).

CP from the CTD

The CP visualization based on the CTD can be seen above. There is a clear relationship between fluorescence and CP as seen below.

The next graph shows the same data but is divided differently. The best relationships can be seen for the observations below the 0.1 isolume and located in open water. I think the bottom left panel is the most interesting where we can see the L shape.

Finally, the same data by transect.

Vertical CP profiles (deepest CTD cast)

The next graphic shows CP vertical profiles for some of the deepest stations below 500 meters. Any ideas why there is almost always a bump in the data in the very last portion of the profiles?

CP from the MVP

Bottom profiles

The next graphs show the longitudinal variability of CP at the deepest locations within each transect. The left column shows how the MVP was moving in the water column whereas the right column shows how CP varied along the transect. For example, we can see that the MVP followed the 300 meters line at transect 100.

Particle backscattering coefficient (bbp)

bbp from the hydroscat

These graphs show bbp at six different wavelengths measured by the hydroscat.

Particulate organic carbon (POC)

bbp vs chlorophyll-a fluorescence (Hydroscat)

The next graphs explore the relationships between bbp and chlorophyll-a fluorescence from the Hydroscat device.

The same data divided differently.

Comparing bbp at two wavelengths (Hydroscat)

Scatterplot between bbp measured at 532 nm and 700 nm.

The same data presented differently.

Phytoplankton absorption

Maybe we could use absorption information to get some additional insights.

Vertical profiles

This graph shows the vertical profiles of phytoplankton absorption at 440 nm. It seems that open water stations have subsurface maximum more apparent than the profiles of the ice-covered stations.

Spectral profiles

By looking at the spectral profiles, we can also notice that there are indeed differences in the water column.

Finally, I have averaged the spectral profiles in four categories. I think these can be seen as end-members spectra.

Comparing blue vs red regions

With all the data.

By groups.

blue/red ratio of specific absorption

We can see that there is an increase in the ratio after owd = 0. I am not sure to understand why the ratio is also higher at depth when owd < 0. Is there a shift when the ice is melting?

POC vs CP

The next graphs show the relationships between POC and CP.

bbp vs chla (Hydroscat)

Note that I have used chla fluorescence from the Hydroscat which is given as raw counts. If we decide to go further with this data, I guess we could convert it into actual biomass/stock quantities.

bbp vs cp

The ratio bbp(532)/cp(660) could be considered as a proxy of particle size and composition, increasing when small or inorganic particles become relatively more abundant than large or organic particles (Xing et al. 2014).

So, based on the next graphs, larger particles are more abundant at depth than at the surface. However, I do not see trends in the data.

xxx

This index was calculated as follow:

\[ \frac{\log(\frac{\text{bbp}(532)}{\text{bbp}(700)})}{0.274} \]

Particles in the water column (UVP)

This graph show how many observations (different than 0) there are for each particle class size measured by the UVP.

In this section, we are exploring the size of the particles in the water column using the data from the UVP. Based on a suggestion from Marcel, we categorized the size of particle into three classes:

  1. small (0.102-0.323 mm),
  2. medium (0.323-1.02 mm)
  3. large (1.02-26 mm)

The next table shows the different UPV particle size ranges that were classified into the three end-member classes. Particle biovolume and concentration of the different UVP class sizes were summed up to calculate the total values within each of the three end-member classes.

particle_size_range particle_size_class
102-128 µm particle_class_small
128-161 µm particle_class_small
161-203 µm particle_class_small
203-256 µm particle_class_small
256-323 µm particle_class_small
323-406 µm particle_class_medium
406-512 µm particle_class_medium
512-645 µm particle_class_medium
645-813 µm particle_class_medium
0.813-1.02 mm particle_class_medium
1.02-1.29 mm particle_class_large
1.29-1.63 mm particle_class_large
1.63-2.05 mm particle_class_large
2.05-2.58 mm particle_class_large
2.58-3.25 mm particle_class_large
3.25-4.1 mm particle_class_large
4.1-5.16 mm particle_class_large
5.16-6.5 mm particle_class_large
6.5-8.19 mm particle_class_large
8.19-10.3 mm particle_class_large
10.3-13 mm particle_class_large
13-16.4 mm particle_class_large
16.4-20.6 mm particle_class_large
20.6-26 mm particle_class_large

Particle biovolume and concentration

This simple boxplot shows how particle biovolume and concentration vary as a function of the three-class sizes. The particle concentration decreases with increasing particle size whereas the biovolume is rather stable.

Here, I averaged the particle concentration within the first 20 meters of the water column and plotted the results as a function of OWD.

  1. The small particle size concentration almost monotonically increases.
  2. The medium particle size concentration increases until ODW = 10 and the decreases.
  3. The large particle size concentration increases until ODW = 0 and the decreases.

This is interesting because this suggests that there are some kind of phenology/timing going on and a temporal shift between the different particle size classes (at least at the surface).

3D plots

Vertical profiles

Here, I have divided the water column into 5 bins and look at the relationships between the concentration and the biovolume of particle.

  • The relationships degrade as the class size increases.

Apparent visible wavelength (AVW)

In this analysis I used a technique proposed in Vandermeulen et al. (2020).

The location of the AVW effectively represents the balance point around which reflectance data is evenly distributed, or more informally, where a Rrs(λ) spectrum would be perfectly balanced on the tip of a pin if each individual channel held a physical weight proportional to its intensity.

This graph shows all the phytoplankton absorption spectra colored by their calculated AVW value.

Using the above data, I calculated the average AVW by depth. We can see that there is a shift from blue to green along the water column, suggesting that there is a change in the spectral shape of phytoplankton.

If we only keep observations above the isolume, the relationship is even clearer.

Figures

Figure 1

Figure 2A

This graph will be used to show an overview of the physical dynamics in the Baffin Bay. Data from Randelhoff et al. (2019).

Figure 2

  • Contrary to what can observed in Hill et al. (2013) (see Fig. 8), we can see underice sub-surface chla maximum in the Greenedge data.

  • The phytoplankton biomass (chla) under the ice seems to represent an important source of primary production (see Appendix 4).

Figure 2C

For this figure, I have integrated the data in the water column:

  • NO3 was integrated between the surface and hBD.
  • Chlorophyll-a was integrated between the surface and 0.1 isolume.

Figure 3

Figure 4

Rosette pigments were summed as follow:

  • Photoprotection: Zeaxanthin, Diatoxanthin, Diadinoxanthin, Antheraxanthin, Violaxanthin

  • Photosynthetic: Sum 19HF-like, Sum 19BF-like, Peridinin, Fucoxanthin, Chlorophyll c3, Chlorophyll c1+c2+MgDVP, Chlorophyll b, Lutein

Figure 5

  • Comparison between total chla from the CTD vs all the pigment groups defined by Joséphine.

  • Pigment groups:

    • photoprotection = Zeaxanthin + Diatoxanthin + Diadinoxanthin + Antheraxanthin + Violaxanthin
    • photosynthetic_pigments = Sum 19HF-like + Sum 19BF-like + Peridinin + Fucoxanthin + Chlorophyll c3 + Chlorophyll c1+c2+MgDVP + Chlorophyll b + Lutein
    • tchla = Total Chlorophyll a
    • phaeophytin_a = Sum Phaeophytin a
    • phaeophorbide_a = sum Phbd a
    • chlorophyllide_a = sum Chld a

Figure 6

Figure 7

Figure 8

  • In this figure, the total particle count is defined as the sum of all particles under 2.05 mm.

Figure 9

  • When bbp/cp increases, small particles become relatively more abundant than large particles.

  • At high bbp/cp (i.e. higher contribution of small particle to the total pool):

    • There are less particles.
    • Chlorophyll-a concentration is lower.
    • There are less particles in open water than in ice-covered stations.
  • In this figure, the total particle count is defined as the sum of all particles under 2.05 mm.

  • Using all the data (not only <= 100 meters).

    • I using a threshold of 5 meters to discard ctd, uvp and hydroscat measurements that were too far apart.

Appendix

Appendix 1

  • I have calculated the number of open water days using the same SIC data as Achim used in his paper (resolution of 3.125 km). I have used a SIC threshold of 15% that lasted at least 3 consecutive days to determine the melting and freezing of sea ice.

  • On the next plot, we can see that there are not many observations with negative OWD because the MVP was deployed in open water.

Appendix 2

Co-variability of subsurface maximums for bbp, cp and chla. At each OWD, the depth of the maximum value (ex.: chla) was extracted. Note that the depths at which each variable is maximum was calculated on raw data and not the interpolated data from Fig. 02. The points are the raw data and the lines represent values fitted using loess.

Appendix 3

  • It was previously shown that chla/cp was a good indicator of phytoplankton photoacclimation (Ek) and physiology (Fv/Fm).

  • Very weak relations with either Ek or Fv/Fm based on the Greenedge data.

Appendix 4

This figure is quite interesting:

  • We can see that high PP occurred before OWD = 0.

  • The correlation between PP and Chl-a is much higher for ice-covered stations compared to open water stations.

    • Any ideas why in situ PP is not correlated to the phytoplankton biomass for stations located in open water? This is in contradiction with what is shown by Hill et al. (2013).

    • Could it be related to nutrients depletion?

TODO

  • Determine if CP correction (baseline shift) should be performed or not.

  • Check if there is a relation between particle size ratio and bbp/cp ratio.

  • Continue the exploration of nutrients vs pp/chla.

  • Add isolume and hBD legends on 3D plots.

References

Hill, Victoria J., Patricia A. Matrai, Elise Olson, S. Suttles, Mike Steele, L. A. Codispoti, and Richard C. Zimmerman. 2013. “Synthesis of Integrated Primary Production in the Arctic Ocean: II. In Situ and Remotely Sensed Estimates.” Progress in Oceanography 110 (March): 107–25. https://doi.org/10.1016/j.pocean.2012.11.005.
Kheireddine, Malika, and David Antoine. 2014. “Diel Variability of the Beam Attenuation and Backscattering Coefficients in the Northwestern Mediterranean Sea (BOUSSOLE Site).” Journal of Geophysical Research: Oceans 119 (8): 54655482. https://doi.org/10.1002/2014JC010007.
Randelhoff, Achim, Laurent Oziel, Philippe Massicotte, Guislain Bécu, Martí Galí, Léo Lacour, Dany Dumont, et al. 2019. “The Evolution of Light and Vertical Mixing Across a Phytoplankton Ice-Edge Bloom.” Elem Sci Anth 7 (1): 20. https://doi.org/10.1525/elementa.357.
Vandermeulen, Ryan A., Antonio Mannino, Susanne E. Craig, and P. Jeremy Werdell. 2020. “150 Shades of Green: Using the Full Spectrum of Remote Sensing Reflectance to Elucidate Color Shifts in the Ocean.” Remote Sensing of Environment 247 (September): 111900. https://doi.org/10.1016/j.rse.2020.111900.
Xing, Xiaogang, Hervé Claustre, Julia Uitz, Alexandre Mignot, Antoine Poteau, and Haili Wang. 2014. “Seasonal Variations of Bio-Optical Properties and Their Interrelationships Observed by Bio-Argo Floats in the Subpolar North Atlantic.” Journal of Geophysical Research: Oceans 119 (10): 7372–88. https://doi.org/10.1002/2014JC010189.